Difficulty in Using Custom Dataset for PointNet Segmentation

[ItsShriks]

Description:
Hello, I am trying to use PointNet Segmentation to train my own model on a custom dataset. However, I am facing several issues related to the filesystem structure and dataset organization. Since I cannot download the dataset referenced with ShapeNet, I am unsure how to properly format my dataset to work with the existing training pipeline.

Issues I am facing:
1. Dataset Structure:
• The provided code references ShapeNet, but there is no clear explanation on how to adapt it for a different dataset.
• What should be the directory structure and required file formats for custom datasets?
2. Preprocessing & Data Loading:
• How should I preprocess my point cloud data to match the expected input format?
• Are there specific scripts in the repository for converting point clouds into the required format?
3. Training Pipeline:
• Which configuration files or parameters need to be modified for training on a new dataset?
• Are there any dataset-specific dependencies or precomputed features required before training?

Request:
Could you please provide:
• A clear guideline on how to structure and preprocess a custom dataset.
• An example or template dataset format compatible with the existing PointNet segmentation pipeline.
• Any references or documentation that could help in setting up a custom dataset.

I would really appreciate any guidance on this! Thanks in advance.